Three studies tested the hypothesis that anticipated regret (AR) increases consistency between exercise intentions and behaviour. Study 1 employed a longitudinal survey design (N = 384). Measures specified by the theory of planned behaviour, past behaviour, and AR were used to predict self-reported exercise behaviour 2 weeks later. AR moderated the intention-behaviour relationship such that participants were most likely to exercise if they both intended to exercise and anticipated regret if they failed to exercise. Study 2 used an experimental design to examine the effect of focusing on AR prior to reporting intentions (N = 229). Exercise was measured 2 weeks later and the AR-focus manipulation was found to moderate the intention-behaviour relationship in a similar manner to that observed in Study 1. In Study 3 (N = 97), moderation was replicated and was shown to be mediated by the temporal stability of intention.